Re: vfork()

["Angelo Dell'Aera" <buffer@antifork.org>]

On Mon, 3 Feb 2003 09:18:55 +0530
"Rupa" <rupa.ramakrishna@ionidea.com> wrote:

> 
> Hello,
>   I  have a  small doubt  regarding  the working  of vfork()  system
> call. when a call to vfork() is made, the child process created will
> be sharing the  address space of the parent. I  want to know whether
> the  pid   returned  by   the  vfork()  is   same  as  the   pid  of
> parent. Because in case of fork() system call, child process differs
> from the  parent by  pid. I want  to know  regarding pid in  case of
> vfork() system call.

You have to know that vfork()  is really similar to fork() syscall and
so , just like fork(), child  PID is different from parent's one.  The
only difference  is that  they share the  same address  space.  Infact
vfork() was  created few  years ago thinking  about the scenario  of a
process which has to fork  a new process and then 'substitute' address
space with  execve().  While  fork() infact created  a copy  of parent
address space  in child, vfork() didn't  do it. So when  you wanted to
exec  another process the  best thing  to do  was vfork()  + exec**().
Infact, copying  an address  space which will  be soon  substituted is
useless.  Now  situation is  really  different  since  modern OSs  use
Copy-On-Write in forking a process. So at current time, vfork() is not
useful as it was in the past.

As regards CLONE_PID  flag (I read it in another mail)  it's used in a
very limited range of situations. For example, if kernel is running on
SMP you  need an init process  on every CPU  and so in this  case it's
useful to  create them with CLONE_PID  (so all init  process will have
PID 1).

Regards, 
Angelo Dell'Aera
--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive:       http://mail.nl.linux.org/kernelnewbies/
FAQ:           http://kernelnewbies.org/faq/